Search for: All records

Recent discoveries of copious amounts of dust in quiescent galaxies (QGs) at high redshifts (z ≳ 1 − 2) challenge the conventional view that these objects have a negligible interstellar medium (ISM) in proportion to their stellar mass. We made use of theSIMBAhydrodynamic cosmological simulation to explore how dust and cold gas evolve in QGs and are linked to the quenching processes affecting them. We applied a novel method for tracking the changes in the ISM dust abundance across the evolutionary history of QGs identified at 0 < z ≲ 2 in both cluster and field environments. The QGs transition from a diversity of quenching pathways, both rapidly and slowly, and they exhibit a wide range of times that elapsed between the quenching event and cold gas removal (from ∼650 Myr to ∼8 Gyr). Contrary to some claims, we find that quenching modes attributed to the feedback from active galactic nuclei (AGNs) do not affect dust and cold gas within the same timescales. Remarkably, QGs may replenish their dust content in the quenched phase primarily due to internal processes and marginally by external factors such as minor mergers. Prolonged grain growth on gas-phase metals appears to be the key mechanism for dust re-formation, which is effective within ∼100 Myr after the quenching event and rapidly increases the dust-to-gas mass ratio in QGs above the standard values (δ_DGR ≳ 1/100). Consequently, despite heavily depleted cold gas reservoirs, roughly half of QGs maintain little evolution of their ISM dust with stellar age within the first 2 Gyr following the quenching. Overall, we predict that relatively dusty QGs (M_dust/M_⋆ ≳ 10⁻³ − 10⁻⁴) arise from both fast and slow quenchers, and they are prevalent in quenched systems of intermediate and low stellar masses (9 < log(M_⋆/M_⊙) < 10.5). This strong prediction poses an immediate quest for observational synergy between, for example, theJames WebbSpace Telescope (JWST) and the Atacama Large Millimetre Array (ALMA). 

A new matrix framework is presented in this studyfor the improved ionization efficiency of complex mixtures by matrix-assisted laser desorption ionization (MALDI) mass spectrometry/imaging. Five nitro indole (NI) derivatives [3-methyl-4-nitro-1H-indole (3,4-MNI), 3-methyl-6-nitro-1H-indole(3,6-MNI), 2,3-dimethyl-4-nitro-1H-indole (2,3,4-DMNI), 2,3-dimethyl-6-nitro-1H-indole (2,3,6-DMNI), and 4-nitro-1H-indole(4-NI)] were synthesized and shown to produce both positive and negative ions with a broad class of analytes as MALDI matrices. NI matrices were compared to several common matrices, such as 2,5-dihydroxybenzoic acid (DHB), alpha-cyano-4-hydroxylcinnamicacid (CHCA), sinapinic acid (SA), 1,5-diaminonaphthelene (1,5-DAN), and 9-aminoacridine (9-AA), for the analysis of lipid, peptide, protein, glycan, and perfluorooctanesulfonic acid (PFOS) compounds. 3,4-MNI demonstrated the best performance among the NI matrices. This matrix resulted in reduced ion suppression and better detection sensitivity for complex mixtures, for example, egg lipids/milk proteins/PFOS in tap water, while 2,3,6-DMNI was the best matrix for blueberry tissue imaging. Several important aspects of this work are reported: (1) dual-polarity ion production with NI matrices and complex mixtures; (2) quantitative analysis of PFOS with a LOQ of 0.5 ppb in tap water and 0.05 ppb in MQ water (without solid phase extraction enrichment), with accuracy and precision within 5%; (3) MALDI imaging with 2,3,6-DMNI as a matrix for plant metabolite/lipid identification with ionization enhancement in the negative ion mode m/z 600−900 region; and (4) development of a thin film deposition under/above tissue method for MALDI imaging with a vacuum sublimation matrix on a high-vacuum MALDI instrument. 

Deep learning based PET image reconstruction methods have achieved promising results recently. However, most of these methods follow a supervised learning paradigm, which rely heavily on the availability of high-quality training labels. In particular, the long scanning time required and high radiation exposure associated with PET scans make obtaining these labels impractical. In this paper, we propose a dual-domain unsupervised PET image reconstruction method based on learned descent algorithm, which reconstructs high-quality PET images from sinograms without the need for image labels. Specifically, we unroll the proximal gradient method with a learnable norm for PET image reconstruction problem. The training is unsupervised, using measurement domain loss based on deep image prior as well as image domain loss based on rotation equivariance property. The experimental results demonstrate the superior performance of proposed method compared with maximum-likelihood expectation-maximization (MLEM), total-variation regularized EM (EM-TV) and deep image prior based method (DIP).